v1.0.1 / chapter 11 of 20 / 01 sep 07 / greg goebel / public domain
* "Superconductivity" is the state in which a material has literally no resistance to electrical current. The phenomenon was discovered early in the 20th century, but for most of the following decades it remained little more than a curiosity. The materials that exhibited superconductive behavior only did so if they were cooled to within a few degrees of absolute zero, which limited their use to highly specialized applications.
Interest in superconductivity skyrocketed in the late 1980s when materials were discovered that remained superconductive at relatively high temperatures, but after the initial excitement wore off, development of practical applications proved painfully slow. However, by the end of the century, work towards applications of superconductive materials in power electric systems, sensors, and digital electronics finally seemed to be on track. This chapter provides an overview of superconductive principles, materials, and applications.
* Superconductivity was discovered in 1911 by Dutch physicist Heike Kammerlingh Onnes (1853:1926). He discovered that when mercury was cooled by liquid helium to 4 degrees Kelvin, it lost all resistance to electrical current. Onnes would later win the Nobel prize for this work. Later research showed that many metals, such as tin, lead, and niobium, were also superconductive when cooled to extremely low temperatures. Given the difficulties of working at such "cryogenic" temperatures, superconductivity remained interesting but of little practical use, though materials were found that became superconductive at slightly higher temperatures. Theoreticians were fascinated by the phenomenon because nobody had any clear idea of why it occurred.
The theoretical principles of superconductivity were finally outlined in 1957, when John Bardeen (one of the inventors of the transistor), Leon N. Cooper (born 1930), and J. Robert Schrieffer (born 1931) published a theory that would also win a Nobel prize. The "Bardeen-Cooper-Schrieffer (BCS)" theory suggested that cryogenic cooling of materials such as niobium suppressed the random thermal noise in their crystal structure. This allowed "phonons", the quantized mechanical vibrations of the crystal lattice, to set up a weak electrical interaction that coupled electrons with opposite spin and momentum together in "Cooper pairs".
This effectively took two half-spin fermions, which obeyed the exclusion principle and so could not occupy the same energy state, into an integer-spin boson, where energy states could be shared. This is one of the interesting differences between fermions and bosons: a system with an even number of half-spin fermions will act like a boson, since two half spins can add up to a whole spin, but a system of bosons can never act like a fermion, since there is no way a sum of integer spins can add up to produce a half spin.
Electrical resistance is caused by the scattering of electrons due to defects, impurities, and thermal vibrations in the crystal lattice of a conductor. However, the binding of electrons in Cooper pairs eliminates scattering, and so electrical resistance disappears. Above a specific "Curie temperature (Tc)", thermal vibrations disrupt the Cooper pairs, and the material becomes resistive again. Intense magnetic fields and high currents can also disrupt the pairs and destroy superconductivity.
* Despite the development of BCS theory, doing anything useful with superconductors remained an uphill struggle. What seemed to be a breakthrough finally occurred in the 1980s. In September 1986, Alex Mueller (born 1927) and Georg Bednorz (born 1950), two scientists at an IBM research center in Zurich, Switzerland, published a paper describing a copper-oxide compound that exhibited superconductivity at 35 degrees Kelvin, compared to 23 degrees for niobium alloys, which had the highest Curie temperature known up to that time. They published their paper in an obscure German physics journal in hopes that it wouldn't be noticed. This tactic allowed them to reinforce their preliminary research without interference, but still be able to prove the priority of their work if other reports were published.
After more studies, the two scientists became convinced that their findings were correct. Once their discovery became widely known, a flood of new "high temperature superconductor (HTS)" materials were discovered. By December 1986, a material had been discovered with a Tc of 38 K. A year later, in early 1987, one team claimed discovery of a compound, "yttrium barium copper oxide" ("YBCO", pronounced "ibco"), that had a Tc of 93 K. This moved the Curie temperatures of superconducting materials from the range of liquid helium temperatures to those of liquid nitrogen temperatures. The reduction in cooling requirements promised to greatly reduce the cost of superconducting technology and widen its range of applications.
The enthusiasm of researchers in the field was manifested that year by a special meeting of the American Physical Society in the Hilton Hotel in New York City, crammed with 3,000 physicists, many of whom stayed up all night discussing the new superconductors. The event became known as the "Woodstock of physics".
* Since 1986, over 100 HTS materials have been discovered. The record Tc now stands at 138 degrees Kelvin. This progress has been made even though nobody really knows how high-temperature superconductivity works.
While there is clearly some electron pairing mechanism involved, as is the case with the old "low-temperature superconductors (LTS)", the phonon-linkage mechanism associated with Cooper pairs in low-temperature superconductors can't work at high temperatures, since thermal vibrations would quickly break the phonon linkages. The most popular theory is that the pair coupling occurs due to subtle magnetic effects created by the HTS lattice, but nobody has been able explain how it happens. There is some confidence that improved studies will be able to close in on the mystery of HTS, but nobody's in a position to say what the ultimate conclusion will be.
Finding a better theory for high-temperature superconductivity is not an academic issue. Understanding what causes the phenomenon will help researchers to address some of the problems they have encountered working with HTS. For example, magnetic vortexes set up by the flow of electrical current through an HTS have a tendency to drift through the material, and this drift dissipates energy, or in other words, causes resistance. The material needs to have strong "flux pinning" to ensure the vortexes do not migrate.
More importantly, a better theoretical understanding may lead to raising the Curie temperature still further. Researchers believe this is perfectly possible, since a copper-oxide compound made with mercury has been shown to superconduct at 164 K when squeezed to extremely high pressure in a diamond anvil. As a result, one avenue of research has been to try to modify superconductive materials into configurations resembling those that they adopt under high pressure.
* After the great expectations raised by the discovery of HTS materials, the enthusiasm of researchers gradually deflated when they found out the practical limitations of HTS. Their excitement at the original discovery of HTS was understandable given that the discovery really was a major breakthrough in materials science, but people became overenthusiastic and predicted widespread short-term applications that simply didn't happen. One remarkably honest researcher who was swept up in the euphoria commented later: "If someone shoves a mike in your face ... you can say some things that, in retrospect, are pretty silly."
The potential did seem to be there. The "critical current density" -- the maximum amount of current a superconductor can support before becoming resistive -- of YBCO is very high, about a million amperes per square centimeter, and the material could remain superconductive in relatively high magnetic fields. Unfortunately, YBCO features irregular grains that are difficult to work into strips and wires, and so attention shifted for a time to another material, "bismuth, strontium, calcium, copper, and oxygen" ("BSCCO", pronounced "bosco" or "bisco") materials. BSCCO has flat, regular grains that can be more easily aligned, and proved much easier to fabricate, although it does not have nearly the current capacity of YBCO.
Researchers have learned to encase grainy, brittle BSCCO material in silver and extrude the assembly into long filaments. The filaments are then rolled and heated to align the BSCCO layers to form a continuous wire. BSCCO has a layered structure; rolling breaks up and spreads the layers, and heating merges them together. The critical current density of BSCCO is about 70,000 amperes per square centimeter. This value is about three times greater than it was in the mid-1990s. Practical BSCCO wires are not made entirely of superconducting material, and so their actual critical current density is about 15,000 amperes per square centimeter.
The silver coating for BSCCO means that it is difficult to reduce prices, so efforts have shifted back towards YBCO, with researchers working hard to deal with the difficulties of working the material. They have learned how to deposit YBCO on tapes of nickel or other metal surfaced with a material such as cerium oxide (CeO) that helps align the YBCO grains. The ribbon is then topped with copper, both to seal in the YBCO and to provide an alternate conducting path if cooling is lost. YBCO has a critical current density of up to 3 million amperes per square centimeter, and so this thin layer can carry huge currents. Progress has been made and there is some optimism that YBCO may win the superconductor race after all.
* A promising new superconductor material, "magnesium diboride (MgB2)", was discovered in 2001 by a team of Japanese materials researchers. Although it was first thought to be an HTS, closer examination showed that its behavior was that of an LTS that had an unusually high Tc of 39 degrees Kelvin. Although that wasn't close to the Curie temperature of modern HST materials, MgB2 was in the ballpark with early HTS materials, and had many good properties: it is cheap, easy to fabricate, and much easier to work into wires than other HTS materials. The difficulty of cooling rises sharply towards infinity as absolute zero is approached, so even the relatively low Tc of MgB2 is still a big deal, permitting MgB2 systems to be chilled with a mechanical cryocooler system, not liquid helium.
MgB2 originally suffered from a low critical current density of about 35,000 amperes per square centimeter and poor resistance to magnetic fields, but researchers are making progress in both these areas. For example, replacing about 5% of the boron with carbon doubled the resistance to magnetic fields.
* Practical applications of superconductors have focused in three areas: electric power systems and devices, such as power transmission lines, electric motors, and transformers; sensitive "superconducting quantum interference device (SQUID)" sensors; and ultrafast superconducting digital logic components.
Electric power systems offer the greatest potential in the near term, and indeed are the primary application for HTS. As it turned out, HTS has some limitations that restrict its usefulness for SQUIDS and logic devices, but substantial improvements have also been made in traditional LTS technology to advance those fields. These applications are discussed below.
* Researchers are now demonstrating full-scale prototypes of electric power cables, motors, transformers, and other heavy electrical gear made with HTS wire. These prototype systems waste much less energy than existing technology; are generally smaller, lighter, safer; and in some cases are more environmentally benign. The US Department of Energy (DOE) has been the primary funding organization for this work in the US.
The DOE estimates that 7% of all electrical power generated in the United States is wasted due to resistive losses. Converting the US power grid to superconducting cables would cut those losses in half. Superconducting power distribution cables have been installed on a trial basis, and have proven lighter and of course much less wasteful than traditional distribution lines. The high cost of the superconducting material makes power trunks using it economically impractical at present, but the costs are dropping, and companies working with YBCO wire believe it will be competitive with copper in a few years.
Experimental power distribution transformers have been built as well. Conventional heavy power transformers are cooled in a bath of "polychlorinated biphenyls (PCBs)", which can catch fire under extreme circumstances, and are also toxic pollutants. A superconducting power transformer does not need to be stored in an oil batch, since it dissipates little power and is stored in liquid nitrogen anyway.
DOE estimates that high-power electric motors use about 30% of US electric power production. DOE has worked with commercial firms to develop an experimental 750 kW (1,000 HP) electric motor using BSCCO wires. The experimental superconducting motor is about half the size and weight of a conventional electric motor with the same output power. It has about the same efficiency as a conventional motor at present, but the development team believes they can cut losses in half, and are working on a motor five times more powerful.
The US Navy is interested in even bigger HTS motors to drive all-electric ships. Conventional electric motors of this size do exist but are too big for naval use. An HTS motor for ship propulsion would be only about a third as big as a conventional motor, and would weigh about 18,000 kilograms (40,000 pounds).
* The theoretical basis for the SQUID was discovered in 1962 by Brian D. Josephson (born 1940), then a research student at the University of Cambridge, who also was awarded a Nobel prize for his work. (It seems likely that the connection between significant discoveries in superconductivity and Nobel prizes has not gone unnoticed by researchers in the field.) Josephson's work was focused on what would happen if two layers of superconductors were separated by a thin layer of insulating material. Such a structure would later be known as a "Josephson junction".
In the superconductors on both sides of a Josephson junction, each of the Cooper pairs act as a single boson, defined by a single quantum-mechanical wave function. The wave functions do not go immediately to zero at the insulating barrier, however, and if the barrier is thin enough, wave functions overlap, forming a single, continuous wave function that allows Cooper pairs to tunnel through the junction. This tunneling makes the insulating barrier a weak superconductor, with the maximum or "critical" current dependent on the temperature, the material, and the size of the junction. This "direct-current" Josephson effect was quickly verified, and a similar "alternating current" Josephson effect was discovered soon afterwards.
SQUID operation is based on quantum-mechanical "flux quantization". Quantum theory dictates that electrons in atoms and molecules can only take on certain discrete energy levels. Similarly, the magnetic field in a superconducting ring can only have certain discrete levels of magnetic field strength, or flux density, though the steps between levels are extremely small. Each small increment defines a "flux quantum" or "fluxon".
A DC SQUID consists merely of two Josephson junctions joined in a ring. The SQUID normally carries a superconducting "bias" current. An external magnetic field applied to the ring will affect this current. The relationship between the external field and the current through the loop is not intuitive. The current through the loop is maximum when the external field is an integer multiple of the flux quantum through the ring, and is minimum when the external field (which, being unconstrained, can take on continuous values) is a half-integer multiple of the flux quanta within the ring. This effect allows measurement of the external field by measurement of the current through the ring, or in practice, the voltage across it. In short, the SQUID acts as an extremely sensitive magnetic field detector.
* The earliest SQUIDs were made by hand in the mid-1960s. Modern SQUID sensors are built using fabrication techniques refined for building integrated circuits. Such mass-produced SQUIDs are implemented as a square ring of niobium bisected by two Josephson junctions, with the insulating junction consisting of aluminum oxide. Hundreds of such SQUIDs can be built on a single silicon wafer, which is then diced to yield individual sensors.
SQUIDs based on AC Josephson effect have also been built. They are known as RF SQUIDs, because they are biased by a current oscillating in the radio-frequency range at about a megahertz. RF SQUIDs consist of a superconducting loop with a single Josephson junction. The loop is coupled to a resonant circuit consisting of an inductor and capacitor in parallel. The amplitude of the RF voltage across the resonant circuit will change in response to an external magnetic field. RF SQUIDs are not as sensitive as DC SQUIDs, but they were originally cheaper to make. Since DC SQUIDs have become easier to fabricate, the use of RF SQUIDs has faded.
To be useful, the output of a SQUID sensor has to be amplified and coupled to a measurement system. Amplification can be performed with a "flux transformer", which is just a larger loop of superconducting material coupled to the SQUID. An external magnetic field induces a supercurrent in the flux transformer, which in turn induces a flux in the SQUID. Such a device can have a magnetic field resolution of one femtotesla (10^-15 tesla), or about 10^-11 the strength of the Earth's magnetic field. SQUID sensors can also be constructed in the form of "gradiometers", devices to measure magnetic field gradients, by using two or more magnetometers coupled to a differential amplifier system, or by a single SQUID with twin flux transformers, wound in opposite directions.
Magnetic-field gradient measurements are particularly useful for medical diagnosis. Magnetic fields due to biological activity are very small, ranging from a few femtoteslas for the brain to about 50,000 femtoteslas for the heart. However, SQUID gradiometers can measure these fields while rejecting stronger remote sources from electric motors and cars. Even though these remote fields are much stronger than the biological fields, their gradients flatten out at long range, and so they are ignored by the gradiometer.
* SQUID sensor arrays, the latest versions of which incorporate hundreds of SQUID sensors, can be used to generate a "contour map" of a magnetic field that helps determine the internal structure of the target and its behavior. Such arrays have been used for a wide variety of purposes, for example the treatment of focal epilepsy, and "cardiac arrythmia", or erratic heartbeat.
Focal epilepsy involves a localized electrical discharge in the brain that causes a seizure, possibly corresponding to scar tissue or some other defect in the brain. SQUID arrays can be used to help pinpoint the site through the magnetic field caused by the discharge. A gamma-ray beam can then be used to destroy the defect. Erratic heartbeat is caused by spurious electrical pathways in the heart that disrupt normal cardiac rhythyms. Finding these pathways by invasive means is difficult, involving probes of the heart with catheters, but SQUID arrays have shown that they can localize the anomaly with much less trouble. The false electrical path can then be destroyed with an electrical discharge from a catheter.
The expense of SQUID arrays has limited their use, but their greater speed and accuracy in comparison to alternate methods, as well as their noninvasive operation, promise to greatly reduce costs of treatment and rehabilitation of conditions such as focal epilepsy and cardiac arrythmia.
* SQUIDS are also useful to perform fundamental physical measurements, ranging from simple measurements of magnetic fields in material samples, to detecting the tiny displacements of large masses used in gravitational-wave detectors.
Such extremely precise measurements require SQUIDs made of LTS materials, but for less demanding applications, cheaper SQUID systems based on HTS materials are now available, though HTS SQUIDS suffer from degraded resolution and higher levels of thermal noise. Devices made of ceramic HTS materials are harder to manufacture than those made of low-temperature superconductors, but fabrication techniques are now available for commercial production of such devices.
HTS SQUIDs have found application in geophysics research. Low frequency electromagnetic fields generated by solar particles falling into the upper atmosphere can propagate into the ground, and SQUID magnetometers, used in conjunction with electrodes buried in the earth, can be used to map underground geological structures. Experiments have also been performed to use SQUID sensors for nondestructive testing, for example, to check for corrosion or other defects in aircraft by examining the magnetic response of currents induced in the skin.
Another interesting new technology based on high-temperature SQUIDs is the "scanning SQUID microscope", which is related to the "atomic force microscope" used to investigate atomic structures of surfaces. Both devices use a tiny probe that is moved by precision control hardware over a surface at microscopic distances, but the scanning SQUID microscope senses magnetic field variations, while the atomic force microscope measures the current tunneling across an air gap into the surface under study.
* Josephson junctions have other useful properties. For example, they switch rapidly from the superconductive to the resistive state, making them attractive as fast logic switching elements ("gates") for high-speed digital logic systems, such as supercomputers. Another one of their useful properties is very low power dissipation, which is just as important for building a high-speed supercomputer.
Above a certain clock rate, the performance of a computer central processing unit (CPU) is limited by the speed of light. The CPU has to access data from memory, and the electrical signal carrying that data can propagate no faster than the speed of light. This means that a superfast CPU has to be very small, but it must also contain a large number of logic gates. If each of these gates dissipates a significant amount of power, then a large amount of power has to be removed from a small volume and cooling becomes a problem. In the extreme case, such a CPU might melt the first time it was turned on.
The low "speed-power product" associated with Josephson junction logic, an order of magnitude smaller than that of any other well-known type of computer logic, provides a way around such obstacles. The latest generation of Josephson junction logic components can operate at switching speeds of hundreds of gigahertz or more, with circuits containing a hundred thousand devices dissipating less than a quarter of a watt.
* There are two classes of Josephson junction logic circuits, the first generation "Josephson latching" logic circuits and the second generation "rapid single flux quantum (RSFQ)" logic circuits. Both classes of devices have been implemented with LTS materials instead of HTS materials, since the power dissipation of HTS devices is about a order of magnitude greater than for LTS devices.
Work on Josephson latching logic began in the late 1960s, with IBM initiating a long-term development program in 1969. Josephson latching logic gates indicated logic states using voltage levels, with a "logic 0" given by zero volts and a "logic 1" given by 3 millivolts. Unlike conventional semiconductor logic gates, Josephson latching logic gates used a pulsed power supply, referred to as a "bias current", operating at the system clock frequency. As their name implied, they also "latched" into their output logic state, that is, they retained their new output even when the input went away, until reset by the next clock pulse.
Josephson latching logic circuits were fabricated using niobium layers separated by aluminum oxide, laid down on a silicon substrate with sputtering. Molybdenum was used for resistors, and amorphous silicon dioxide for insulation. Building logic components was only part of the battle, however. There were also the problems of connecting the three-millivolt logic levels of the superconducting circuitry to conventional logic in the non-cryogenic parts of the computer, and of thermally isolating the superconducting CPU from the hot outside world, while still providing connection to low speed devices like disks and I/O interfaces. Thermal cycling -- the change in temperature from when the system was at room temperature and when it was cooled for operation -- was another major challenge.
The engineering challenges were so great that IBM abandoned their work on superconducting computers in 1983. As it turned out, the discovery of HTS proved to have little or no impact on superconducting logic, and the defense drawdown of the 1990s following the end of the Cold War resulted in a violent shakeout among supercomputer vendors competing for US government dollars.
* Just as the first generation of superconducting logic devices were dying out, a team of Soviet researchers at Moscow State University were developing the first RSFQ logic devices. RSFQ logic used the presence or absence of a single fluxon to indicate a zero or one, with the presence of a fluxon measured as a voltage. RSFQ logic promised better performance than Josephson latching logic and generally seemed a much more promising avenue of investigation.
In 1991, the Russian researchers emigrated in a group to the US, where they have continued their work on RFSQ logic and have tried to bring it into commercial use. As with Josephson latching logic devices, RSFQ logic are constructed using integrated circuit fabrication schemes to lay down layers of niobium and aluminum oxide, with molybdenum resistive elements. However, fabrication techniques have been steadily improved. Researchers believe that they can scale down current RSFQ logic devices by a factor of ten, increasing their performance by a factor of ten as well.
While RSFQ logic devices require the use of cryogenic refrigeration gear, new cryocoolers have been developed that are much cheaper, more powerful, and more reliable than earlier refrigeration systems. A refrigeration unit now costs less than $20,000 USD and can fit into a standard equipment rack.
RSFQ logic devices are projected to initially be used in applications such as high-speed analog to digital converters, with applications in RF communications systems. Such small-scale applications will likely be followed by superconducting CPUs for the next generation of supercomputers.
* Along with superconducting digital logic systems, work has been performed on superconducting imaging sensors -- or more generally "photon detectors", since they don't necessarily operate in the visible light region. Some superconductive sensors are capable of detecting single photons. There are two types of superconducting photon sensors, including "thermal detectors" and "pair-breaking detectors", both based on LTS.
There are various schemes for thermal detectors, but the most common uses a "picture element" or "pixel" made of a superconductive material in a simple DC circuit, in series with a small coil, which is magnetically coupled to a SQUID loop. The superconducting pixel is held right at its Tc; when a photon falls on the pixel, the rise in temperature increases the resistance of the pixel, and so the current in the circuit falls. The drop in current causes the coil to generate a magnetic pulse, which is picked up by the SQUID.
The idea had been around since the 1940s, but it proved difficult to get to work. The trick was finally to maintain a constant voltage across the pixels, with the current providing the heating. Any changes in the resistance of the pixel will reduce the current and heating, bringing the pixel back down to its threshold -- though the momentary change in current will still be picked up by SQUID.
The pair-breaking detector looks like a Josephson junction set up as a pixel. A small bias voltage is applied across the pixel, but a small magnetic field is applied to the insulating barrier to keep Cooper pairs from crossing. The pixel is kept at a low enough temperature to keep it superconductive at all times. When the pixel is hit by a photon, it disrupts Cooper pairs, releasing free electrons -- referred to as "quasiparticles" in this context -- which can pass through the insulating junction and are immediately swept out as a measurable current.
For both classes of devices, the size of the current is proportional to the energy (and so frequency) of the incident photon, and so both can not merely detect photons but can determine their wavelength. They can both operate from the microwave wavelengths to gamma rays. While individual sensors can be put to good use, more often they are arranged as arrays, with rows and columns of pixels. Such an array architecture implies that there is a single electrical conversion circuit for each "column" of the array, with the "rows" of the array switched or "multiplexed" to allow them to be read out in turn.
The superconducting photon sensors have been used in radiation counters whose spectroscopic capabilities allow them to identify particular controlled radioactive materials and not be confounded by background radiation. Arrays are also useful in astronomy, particularly in the millimeter and submillimeter regions of the electromagnetic spectrum. A wide range of other applications is also being investigated.
